[PATCH] msvcrt: Fix reading BOM-less files opened with ccs=unicode

[Martin Storsjo]

This fixes a regression in running MSVC 2010 in wine, when reading
.def files (regressed in b0ab1b76029eaface724a28483fe8b0ea36029dc).

Signed-off-by: Martin Storsjo 
---
 dlls/msvcrt/file.c       |  1 +
 dlls/msvcrt/tests/file.c | 37 +++++++++++++++++++++++++++++++++++++
 2 files changed, 38 insertions(+)

diff --git a/dlls/msvcrt/file.c b/dlls/msvcrt/file.c
index ba534568119..258ae2ac811 100644
--- a/dlls/msvcrt/file.c
+++ b/dlls/msvcrt/file.c
@@ -2184,6 +2184,7 @@ static int check_bom(HANDLE h, int oflags, BOOL seek)
         oflags = (oflags & ~(_O_WTEXT | _O_U8TEXT)) | _O_U16TEXT;
     }else if (seek) {
         SetFilePointer(h, 0, NULL, FILE_BEGIN);
+        oflags = oflags & ~_O_WTEXT;
     }

return oflags;
diff --git a/dlls/msvcrt/tests/file.c b/dlls/msvcrt/tests/file.c
index ecf99e38b0e..b094ca5d1e8 100644
--- a/dlls/msvcrt/tests/file.c
+++ b/dlls/msvcrt/tests/file.c
@@ -1039,6 +1039,43 @@ static void test_fgetwc_unicode(void)
     ch = fgetwc(tempfh);
     ok(ch == WEOF, "got %04hx, expected WEOF (unicode)\n", ch);
     fclose(tempfh);
+
+    // Test BOM-less utf8
+    tempfh = fopen(tempfile, "wb");
+    ok(tempfh != NULL, "can't open tempfile\n");
+    ret = WideCharToMultiByte(CP_UTF8, 0, wchar_text + 1, ARRAY_SIZE(wchar_text) - 1,
+                              utf8_text, sizeof(utf8_text), NULL, NULL);
+    utf8_text[ret] = '\0';
+    ok(ret > 0, "utf-8 conversion failed\n");
+    fwrite(utf8_text, sizeof(char), ret, tempfh);
+    fclose(tempfh);
+
+    // ccs=utf-8 interprets BOM-less content correctly
+    tempfh = fopen(tempfile, "rt, ccs=UTF-8");
+    ok(tempfh != NULL, "can't open tempfile\n");
+    for (i = 1; i < ARRAY_SIZE(wchar_text); i++)
+    {
+        ch = fgetwc(tempfh);
+        ok(ch == wchar_text[i],
+           "got %04hx, expected %04x (utf8[%d])\n", ch, wchar_text[i], i-1);
+    }
+    ch = fgetwc(tempfh);
+    ok(ch == WEOF, "got %04hx, expected WEOF (utf8)\n", ch);
+    fclose(tempfh);
+
+    // ccs=unicode without a BOM only returns raw bytes
+    tempfh = fopen(tempfile, "rt, ccs=unicode");
+    ok(tempfh != NULL, "can't open tempfile\n");
+    for (i = 0; utf8_text[i]; i++)
+    {
+        ch = fgetwc(tempfh);
+        ok(ch == (unsigned char) utf8_text[i],
+           "got %04hx, expected %04x (unicode[%d])\n", ch, (unsigned char)utf8_text[i], i);
+    }
+    ch = fgetwc(tempfh);
+    ok(ch == WEOF, "got %04hx, expected WEOF (unicode)\n", ch);
+    fclose(tempfh);
+
     unlink(temppath);
 }

[Piotr Caban]

Hi Martin,

I've sent a changed version of your patch. I hope it looks good for you.

Thanks,
Piotr

[Martin Storsjö]

Hi Piotr,

On Wed, 27 Jan 2021, Piotr Caban wrote:
> I've sent a changed version of your patch. I hope it looks good for you.

Thanks! Sorry I always forget about not using C99 style comments - I don't 
mind them removed if you (who most probably is the one going to poke 
around in the test later) don't want them littering the test code. For the 
functional change I don't have much input - if it fixes the testcase I'm 
happy at least :-)

// Martin